Think Of Virtualization Like A Mutual Fund
John McKnight gave me this analogy, and I can't be happier about it. It really helps people understand what this "virtualization" thing is all about.
You used to pick stocks. You watched them perform, you bought and sold based on that performance, and you always tried to optimize the outcome. You most likely also did this as a self-created part-time job, since you probably were paid to do something else. Regardless, you did it - and it was almost feasible when you had a few stocks to watch, but once you had dozens or hundreds or thousands, it becomes an exercise in futility.
Most people with real jobs don't do that as much anymore. They buy mutual funds. They can create objectives for their overall performance requirements and have a single view into the fund in order to make macro decisions. People don't have to worry about watching all 100 stocks in the fund - because someone else does it - with the help of automation. In our virtual view, we really just judge the fund manager. We don't look at the individual components (stocks), we view it holistically. Imagine if every time the fund manager wanted to make a change - for all the right reasons, they had to call us and ask us permission. How inefficient that would be. What if every time a change was made to one stock, it kept me the individual from performing my job until that change was complete? That would get you fired I think.
Infrastructure virtualization is the mutual fund. Instead of stocks think about file servers or block devices. There are tons of them out there, each one managed individually. How inefficient. By virtualizing the file and/or block world, we can put a portal up to the user. When we need to move their data from one box to another (for good reason or bad) - today we stop them from doing their job. By creating an abstraction layer the user doesn't need to be affected when we make the move - just like when we sold or bought an individual stock within our mutual fund - we don't know it. All we care about is that we are up, we can work, and the performance is acceptable. The IT manager (or the fund manager) is judged on the overall availability and performance of the holistic shop - not one box, just like we wouldn't judge the fund manager on the performance of one stock.
Applying technology behind the scenes to automate processes in order to optimize the overall performance and economic return on the fund is not different than trying to do the same for the IT infrastructure. Keeping the moves from affecting the productivity of the "user" is always a good idea. Fund managers use computers to track the individual stock assets and apply intelligence to make decisions on them - i.e. if the price hits X then sell, Y then buy, etc. - all within the holistic portfolio's mission. This is like ILM for a mutual fund. In our world, once we obviated the need to screw up the individual users life by micro managing infrastructure, wouldn't it be great if we applied automated intelligence to that infrastructure behind the scenes? Our policy might be "if a file gets hit X number of times in this time frame, move it from A to B or replicate it to D - and once it cools off, put it back" or "if we haven't accessed this file for Y period, move it from expensive asset 1 to cheapo asset 4". You get the point.
Perhaps the biggest thing to take out of this train of thought is that it is inevitable. There were lots of people who made a fortune on Wall St. doing things "the old fashioned way" who fought the advancement of technology. They lost. You can still design a car by hand, but no one does. You can still not use electronic banking, but why? Would you use an accountant that didn't use a computer today? Change scares people, but change for efficiency improvements cannot be stopped.
It would be foolish continue to do things the same way in IT even if we weren't adding new stuff constantly. IT Infrastructure Virtualization will happen because it must happen. Sure, there will be some folks left behind, clinging to their abacus's, but that's the price of progress.
Steve,
Thanks for thought!
I am a big fan of total abstraction, or virtualize, for everything... Back when I was in grad school I invented a virtual disk spread out across an entire network. Using TCP/IP and a few UNIX boxes. This was before the internet took off about 1990. Nobody understood it, and I did not know how to market new concepts then, and perhaps the technology to make it commercial was not there yet. But that is the starting point, why does anybody care that they have a disk drive or a flash or even how much memory they have...
I commend you for thinking this way. There will always be resistence to giving up control of lower
infrastructure. And a slew of justifications why it can't happen.
Right now I do the same thing with bandwidth optimization on a Network. Why do adminstrators feel the need to set bandwidth , control and QOS policy down to the bit? They are bright people but they can't make the conceptual jump that there might be a better way to get the same business results much like your stock market analogy.
Thanks Again
Art Reisman
www.netequalizer.com
Posted by: art reisman | June 19, 2006 at 03:33 PM